High sampling rate thermistor string observations at the slope of …tmp.hansvanharen.nl/pedef/os-1-17.pdf · 2017-01-11 · the R-parts of a certain type of RC-oscillator, a Wien(bridge)

Ocean Science, 1, 17–28, 2005www.ocean-science.net/os/1/17/SRef-ID: 1812-0792/os/2005-1-17European Geosciences Union

Ocean Science

High sampling rate thermistor string observations at the slope ofGreat Meteor Seamount

H. van Haren1, R. Groenewegen1, M. Laan1, and B. Koster1

1Royal Netherlands Institute for Sea Research (NIOZ), P.O. Box 59, 1790 AB Den Burg, The Netherlands

Received: 15 November 2004 – Published in Ocean Science Discussions: 22 December 2004Revised: 7 March 2005 – Accepted: 30 March 2005 – Published: 10 May 2005

Abstract. A high sampling rate (1 Hz) thermistor string hasbeen built to accommodate the scientific need to accuratelymonitor high-frequency and vigorous internal wave and over-turning processes in the ocean. The thermistors and theircustom designed electronics can register temperature at anestimated precision of about 0.001◦C with a response timefaster than 0.25 s down to depths of 6000 m. With a quickin situ calibration using SBE 911 CTD an absolute accuracyof 0.005◦C is obtained. The present string holds 128 sen-sors at 0.5 m intervals, which are all read-out within 0.5 s.When sampling at 1 Hz, the batteries and the memory capac-ity of the recorder allow for deployments of up to 2 weeks.In this paper, the instrument is described in some detail.Its performance is illustrated with examples from the firstmoored observations, which show Kelvin-Helmholtz over-turning and very high-frequency (Doppler-shifted) internalwaves besides occasionally large turbulent bores moving upthe sloping side of Great Meteor Seamount, Canary Basin,North-Atlantic Ocean.

1 Introduction

After successful deployments of the first NIOZ thermistorstring (NIOZ1: 32 sensors at 1 m intervals, sampled once per20 s, precision better than 4·10−5◦

C; van Haren et al., 2001)plans were made to modify the concept, so that it would bebetter adapted for different environments. Specifically, wewanted to monitor over periods of a few weeks very fast(∼1 Hz) non-linear motions associated with internal “waves”in “solibore” form above sloping bottoms. Although suchmotions were captured with NIOZ1 (Hosegood et al., 2004),we wanted more detailed information, even more than pre-sented by Thorpe (1987) who used an array of 11 thermis-

Correspondence to:H. van Haren([email protected])

tors at, on average, 10 m spacing, sampling every 10 s to aresolution of 10−4◦

C at∼3000 m depth off Porcupine Bank.Except for Thorpe’s deep-sea measurements (Thorpe,

1987) and in shallow water (Thorpe and Hall, 1974), veryfew “moored” observations have been made of highly vary-ing temperatures as a measure for density (ρ) variations.Such variations can occur as internal gravity waves that aresupported by larger-scale stable density stratification, or asturbulence in mixing events. These processes are relatedbecause internal waves can generate turbulence when theybreak.

The frequency (σ) of free propagating linear internal grav-ity waves is limited tof <σ<N , N�f . At the low end,f =2� sinϕ denotes the local inertial frequency, the normalcomponent of the Earth’s angular momentum vector� mea-

sured at latitudeϕ. At the high end,N=(−gdlnρ/dz)1/2

the background buoyancy frequency, withg the accelerationof gravity, pointing in the downward, negative z-direction.The vertical length scale ofN is crucial for the naturalfrequency of vertical oscillatory motion. However, mo-tions at all scales cause density stratification changes (strain-ing) of many different length and time scales (Pinkel et al.,1991). Thus, internal waves appear not just as linear waves,and, in the time domain, non-linear internal waves showsteep ramps, i.e. sudden temperature changes within a pe-riod of less than 1 min only (Thorpe and Hall, 1974; Thorpe,1987; Gemmrich and van Haren, 2001). Such steep wavesmay overturn when they propagate in larger-scale currentshear thereby generating turbulent mixing through Kelvin-Helmholtz (shear) instability (Turner, 1973). In the openocean typical vertical mixing scales are several meters, unsta-ble overturning remaining for periods of several minutes–1 h,as determined using 3–4 min repeated CTD and microstruc-ture profiles (Alford and Pinkel, 2000).

Even faster temperature variations of O(1 s) having verti-cal scales of O(1 m) have been observed in the upper 100 mof the ocean using “towed” thermistor string observations,

© 2005 Author(s). This work is licensed under a Creative Commons License.

18 H. van Haren et al.: High sampling rate thermistor string observations

Table 1. Specifications of NIOZ thermistor strings. Compared to the very precise NIOZ1 the new NIOZ2 is somewhat less precise but it cansample many more sensors in a much shorter period of time.

NIOZ v.1 NIOZ v.2

Number of sensors 32 (at 1 m intervals)*** 128 (at 0.5 m intervals)***Depth rating 6000 m 6000 mRange (T) −5...55◦C −2...50◦CPrecision <5 · 10−4◦

C 1.5 · 10−3◦C

Thermistors* 2 per sensor 1 per sensorSelf heating ∼=3 · 10−4◦

C ∼=1 · 10−3◦C

Response time (τ ) <0.25 s (in water) <0.25 s (in water)Total sampling time <4 s (for 32 sensors) 0.5 s (for 128 sensors)Minimal sampling interval (1t) 20 s 1 sMemory and battery life 100 days (1t=30 s; 32 sensors) 15 days (1t=1 s; 128 sensors)**

* Siemens Matsushita B57017-K822.** 512 MB Flash card installed. Sufficient power capacity to upgrade to 2 GB.*** Maximum distance between sensors. Lines are flexible to make sensor distances smaller.

0.090.12

M2

02

01

Fig. 1. Mooring site, positions of CTD01 and 02 and local coordi-nate axes at GMS. The M2 tidal ellipse (with major/minor axes in ms−1) is from observations by Mohn and Beckmann (2002). Ampli-tudes are about 1.5 times those from a numerical model they usedand about 3.4 times the far field CB values in the same model.

whilst sampling at 0.1–4 Hz (e.g. Marmorino et al., 1987;Moum et al., 1992; Brandt et al., 1999). These thermis-tor strings had an accuracy of∼10−2◦

C (Marmorino et al.,1987; Selschopp, 1997). As suggested from the “moored”observations by Thorpe (1987) and Hosegood et al. (2004),such temperature variations can be expected even in deep-ocean environments, despite the local background buoyancyperiod, which is relatively large O(10–100 min).

As a result, for monitoring such deep internal tempera-ture variations standard moored thermistor strings sample tooslow, typically once per minute at an accuracy of∼5·10−2◦

C.To monitor the above variations and scales at an accuracy

better than 10−2◦C one should modify a string of the towed

type to one that can be moored. Alternatively, one could con-struct an (expensive) chain of individually logging tempera-ture sensors, such as SBE39. These sensors are quite accu-rate (2·10−3◦

C), but presently they cannot sample faster thanonce per 3 s, whilst running out of power within 6 days whensampling at this rate. So far, a thermistor string has not beenmarketed that meets the above requirements.

Thus, new insights are expected from the newly builtmoored thermistor string “NIOZ2” that can resolve the finestnon-linear internal wave fluctuations and scales of O(1 m)and O(1 s) over a range of 50–100 m sampling at a rate of1 Hz. This string also has the potential, with some limita-tion, to monitor the effects of overturning via eddy diffusivityK estimates from “Thorpe overturning” scalesd, which arefound from re-ordering an unstable density profile to a stableone:K=0.1d2/N (Thorpe, 1987). The limitation to estimatesuchK using NIOZ2 is the present vertical sensor separationof 0.5 m, which is somewhat large, where one would like toresolve O(0.01–0.1 m) scales.

Following the principle of operation of NIOZ2, observa-tions are presented from its first moored deployment near thetop of Great Meteor Seamount (GMS) at latitudeϕ=30.00◦,at whichf equals the diurnal tidal frequencyK1 (Fig. 1).The sloping boundary layer of this site is expected to showinteresting variations, because the dominant tidal currentshave their major axis directed more or less perpendicular tothe slope (Mohn and Beckmann, 2002). These tidal currentsare described as mainy barotropic. The only substantial inter-nal tidal wave motions at GMS have been found at the diurnalfrequencyK1, but to the south of the present mooring.

Ocean Science, 1, 17–28, 2005 www.ocean-science.net/os/1/17/

H. van Haren et al.: High sampling rate thermistor string observations 19

Attachment NIOZ2

ARGOS beacon Acoustic releases

300 kHz ADCP

NIOZ2 elements

sensor

central unit

concentrator

x16

x8

NIOZ2 coiled up on top Rosette frame

Bottom lander

Fig. 2. Mooring configuration and NIOZ2 elements.

2 Instrumentation

2.1 Technical aspects

Comparing NIOZ2 with NIOZ1 several technical differencesare noted (Table 1). NIOZ1 uses two thermistors that are boththe R-parts of a certain type of RC-oscillator, a Wien(bridge)oscillator. This makes the temperature measurement actuallya period measurement. The resolution of such measurementcan be made very high and is independent of cable losses.In contrast, NIOZ2 operates a single thermistor at 0.3 V DCin a Wheatstone bridge on a 24 bit A/D convertor. This re-sults in much less precision (∼10−3◦

C) compared to NIOZ1(∼4·10−5◦

C). However, the power consumption is greatly re-duced and sampling is truly synoptic. The processing time ofall 128 thermistors is only 0.2 s, so that sampling at a rate of1 Hz is easily achieved inclusive of data transport and stor-age, which is 20 times faster than NIOZ1.

In NIOZ2, the glass embedded temperature sensitive semi-conductor plus electronics board are held inside a pressureresistant titanium tube 13 mm in diameter and 120 mm inlength (Fig. 2). The glass embedding ensures no effects ofpressure up to at least 700 Bar on the temperature measure-ment. The advantage of the titanium housing is less corrosionsensitiveness but the disadvantage is the greater difficulty infinding proper casting and primer materials to watertight theconnections to the electric cables. Presently polyurethaneresin is used (85 shore-A). To spread the risk of accidentalleakage the entire string is split up into 8 sections of 16 sen-sors. Each section separately powers 4 groups of 4 sensors.Depending on the type of failure, one either looses data from

1, 4 or 16 sensors at a time. This approach also allows foreasy reconfiguration or replacement in the field. The draw-back is the rather large amount of individual cables involved,which are kept as thin as possible (0.003 m). The 16 cablesfrom the 16 sensors are led to one data concentrator and the 8cables from the 8 data concentrators are led to a single centralcontroller-datalogger-power supply unit.

The data are logged in portions of 10.5 Mb on a 512 Mbcompact flash card using a Persistor CF-1 micro controlleroperating in PicoDos. The uploading of a full 512 Mb mem-ory card takes only about 40 min (200 Kbaud) via an USB-port. As a fresh battery pack outlasts the filling of 512 Mbby a factor of 1.5–2 (depending on the sampling rate), the to-tal period of measurements depends on the capacity of thememory card and on the sampling frequency. The lattercan be chosen between 0.033 and 1 Hz. Although designedfor short-term moorings whilst sampling at high frequency,NIOZ2 can be moored for up to 400 days whilst samplingonce every 30 s.

In NIOZ1 one could replace an individual sensor rathereasily, but no cabling could be changed as it consisted of asingle umbilical. In NIOZ2 one may replace one segment,that is 16 sensors and one concentrator, as the smallest inter-changeable part. Each sensor’s cable length can be modifiedindividually. Presently, all electric cables, sensors and con-centrators are taped to a 0.009 m nylon coated steel cable.The 128 sensors are fixed at 0.5 m distances.

www.ocean-science.net/os/1/17/ Ocean Science, 1, 17–28, 2005


Table 2. Instrument details of ADCP-NIOZ2 mooring between 11March 2003 and 16 March 2003 at 29◦59.876′ N, 028◦18.906′ W,531 m depth. (ls = lowest sensor; fb = first bin).

Instrument depth (m) Sampl. int. (s)

Nortek AquaDopp 445 30Aanderaa RCM11 446 60NIOZ2 528.15 (ls) 1RDI 300kHz ADCP 529.75 (fb: 523.2) 30

2.2 The mooring configuration

The central steel cable assures that the thermistor string canbe moored in line. Generally, the thermistor string is mooredin conjunction with an upward looking 300 kHz,θ=20◦ beamangle RDI-Sentinel ADCP mounted in a bottom landingframe (Fig. 2). Thermistor string motions are kept as lowas possible (<5◦ for currents up to 0.4 m s−1) by attachingthe string under a single ellipsoidal buoy having 200 kg netbuoyancy.

The ADCP measures all three components of current be-tweenz=8–86 m (445–523 m depth). The Cartesian (East,West, Vertical) current components are transferred to (offs-lope, alongslope, bottom normal) = (u, v, w) as in van Harenet al. (1994), withz=0 at the bottom for the thermistor string-ADCP configuration (CTD observations will be referred toas depth from the surface). The ADCP cannot sample fasterthan once per 30 s as this would already fill its entire memoryof 450 Mb within 15 days.

Instead of the ADCP’s tilt meter information, which gavea tilt of 6.7±0.05◦, the bottom slope estimate ofα=4.5±0.3◦

from the ship’s echo sounder is used. Although the ship’secho sounder has a relatively large foot print of∼170 m atthe mooring site, this larger scale information for determin-ing the bottom slope is preferred above the ADCP’s tilt in-formation, because the former length scale is closer to thetypical vertical scale of the boundary layer to be studied (seeSect. 3). Also we have no information on the precise situationof the lander with respect to small bottom structures such asboulders. Given the 4.5◦ bottom slope and an expected localaspect ratio of O(0.01–0.1) the above coordinate transformimplies an O(0.1–1) change inw velocity as the transformfrom originalwraw involvesw ∼wraw · cosα+uraw · sinα.

The ADCP also provides an “error velocity”e (van Harenet al., 1994),

e =b1 + b2 − b3 − b4

4 cosθ=

−(w1 + w2 − w3 − w4)

4+

(u2 − u1 − (v4 − v3)) tanθ

4,

where the sub-scripts “raw” are left out. Thus,e is the differ-ence in beam currentsbi between beam pairs and indicative

of the current inhomogeneity across the beam spread (and/ora failure of one or more beams). As a result, its data can beused to verify the level of significance ofw with respect toinstrumental noise, becausew is defined as,

w =−b1 − b2 − b3 − b4

4 cosθ=

w1 + w2 + w3 + w4

4−

(u2 − u1 + (v4 − v3)) tanθ

4.

It is noted that the system’s output ofe has to be divided by afactor of 4 cosθ as indicated above to be comparable withw

(van Haren et al., 1994). It may be evident thate can also beused to qualitatively investigate quasi-turbulent overturningmotions varying on scales<25 m horizontally and< severalminutes in time.

Finally, the ADCP backscatter strength or echo intensityoutput I is used as a qualitative measure for variations inacoustic reflectances such as suspended material and turbu-lence and small-scale stratification. Instead of using the rawdata that include the water attenuation of sound, and insteadof using a rather imprecise general formula that accounts forthis water attenuation, the mean at each depth is subtracted:the relative echo intensitydI .

As will be apparent from the observations in Sect. 3 theamount of scatterers posed a problem on the data analysis:in general the signal to noise (s/n) ratio was low, presumablydue to clear water. As a result, it was impossible to performspectral analysis and direct estimates of heat and momentumfluxes were unreliable, also because of the large mismatch inresolved scales between NIOZ2 and the present ADCP. Fur-thermore, at some vertical positions like atz=19 m the datawere flawed by the zero-bias due to reflection off NIOZ2,which is attached to the centre of the bottom lander, witha 1 m horizontal off-set with respect to the central verticalaxis of the ADCP (Fig. 2). The latter does “hear” NIOZ2,because this string is more bulky than a standard thermistorstring due to its concentrators and large number of cables.Such unwanted reflections could have been partially avoidedwhen the ADCP were gimballed in the frame. However, theADCP is not gimballed in the frame, because a frame tiltedat a sloping bottom or with one leg at a single boulder maycause one or more beams to reflect at the frame, as the ADCPis mounted inside the frame to prevent entanglement with thethermistor string during deployment.

The bottom lander mooring is recovered using two Ben-thos acoustic releases mounted in the frame, which uncou-ple simultaneously the thermistor string and the main 500 kgdropping weight from the bottom frame, so that the thermis-tor string and the bottom lander surface separately.

3 First results

The bottom lander mooring was located near the top of theeastern slope of GMS, Canary Basin (CB), North-Atlantic



12 12.5 13

−550

−500

−450

T (oC)

−z

(m)

a

35.7 35.8 35.9

sal (psu)

b

0 50 100

−550

−500

−450

N (cpd)

−z

(m)

c

12 12.5 13 13.526.95

27

27.05

27.1

27.15

T (oC)

σ θ (kg

m−

3 )

d

van Haren Figure 3Fig. 3. Environmental conditions from CTD01 (24 Hz raw data;black) and from CTD02 (2 m smoothed; red) (Fig. 1) around thedepth of NIOZ2 (vertical bar in (a) indicates the range).(a) Tem-perature. (b) Salinity. (c) N computed over 5 m intervals.(d)Temperature-density anomaly plot with the corresponding linear re-lationship for CTD01 in green.

ocean (Fig. 3; Table 2). The mooring was deployed for∼5 days due to limited shiptime. The NIOZ2 recorder col-lected 185 Mb of 1 Hz data from about 500 m depth. Themooring was not deployed deeper, because during the pre-deployment calibration the aforementioned casting materialshowed problems at pressures greater than about 200 Bar.Also, ADCP-data become poor when the instrument ismoored at angles of more than 15◦. As GMS is very steepwith typical slopes of 10–20◦ or more, a suitable slope wasonly found near the foot of the underwater mount at∼4500 mdepth and near its top between 450–550 m depth. The ob-served ADCP’s tilt angle of 6.7◦ ensured good current data.The current meter immediately above the thermistor stringshowed typical tilt of 1.5◦, with a maximum of 4.4◦. Asa result, vertical (horizontal) excursions of a NIOZ2 sensorhalfway the string were typically 0.06 (1.5) m, and maximum0.15 (4) m, respectively.

Upon recovery it was found that 3 complete segments ofNIOZ2 had leaked and failed, and that several other sensorshad failed as well. As a result, specific quantitative computa-tions were limited due to this failure of sensors. For example,computation of Thorpe scales was more difficult than antic-ipated given the limited vertical ranges of good data. Nev-ertheless, this first field experiment provided detailed tem-perature observations of internal waves and large non-linearbores.

3.1 Observed environmental interior conditions

The upper ocean of the CB (∼30◦ N, 22◦ W, 5000 m depth)shows a variable rather than a smooth temperature and salin-

−z

(m)

b

73.45 73.5 73.55 73.6 73.65 73.7

−60

−50

−40

−30

−20

−10

yearday

−z

(m)

T (oC)1 min

c

van Haren Figure 4

73.488 73.49 73.492 73.494 73.496 73.498

−50

−40

−30

−20

−10

12.2

12.4

12.6

12.8

13

13.2

69 70 71 72 73 74 75

12

13

14

15

yearday

T (

o C)

a

Fig. 4. (a) Time series of temperature during the entire mooringperiod measured by the ADCP atz=1.25 m and by the AquaDoppcurrent meter moored above NIOZ2 thermistor string atz=86 m.(b)Detail of 10 h of NIOZ2 temperature demonstrating a sharp near-bottom front. The blue bar abovez=52 m indicates two failing seg-ments, the brown bar aroundz=30 m one failing segment; otherwise3 sensors were bad as well. The white rectangle indicates panel(c).Further detail of 17 min of NIOZ2 data around the passage of thefront.

ity change with depth, also in the range of interest here(Fig. 3). This small-scale variability in CTD-profiles sug-gests internal wave induced straining and occasional over-turning. This is observed in profiles obtained above theabyssal CB away from boundaries, as well as for profiles ob-tained above rugged topography like GMS. It is noted thatnear-bottom overturning and stratification vary considerablywith time. Sometimes, stratification moves very close to thebottom, as can be seen with some difficulty in Fig. 3a (redcurve), where it is only 1 m from the lowest level in the pro-file, or about 4 m from the bottom.

The CTD profiles around the depth range of the deployedNIOZ2, demonstrate a slight difference between GMS andthe abyssal CB. This results in a slight change in large-scaletemperature and salinity variation up to 150 m above the bot-tom, or several km horizontally. This affects the “large-scale” stratification. In the open oceanN (computed over5 m) has a value of 45±15 cpd over the depth range in Fig. 3c,although strongly varying between 0–80 cpd due to the smalllayers. As shown in Fig. 3c,N (450 m)=50±20 cpd observedabove GMS, decreasing toN (500)=20±15 cpd and increas-ing again toN (535)=60±20 cpd before becoming negligiblysmall, except near the bottom. This greater variability, alsoat large scales, near GMS may be related to typical slopingbottom processes and internal waves.

The CTD-profiles from different locations around themooring site did not result in great differences intemperature-potential density anomaly (σθ ) relationship.



−z

(m) u

(10−2 m s−1)69.5 70 70.5 71 71.5 72 72.5 73 73.5 74 74.5−80−60−40−20

−20

0

20

−z

(m) v

−80−60−40−20

−20

0

20

−z

(m) w

−80−60−40−20

−2

0

2

−z

(m) e

−80−60−40−20

−2

0

2

yearday

−z

(m) dI

(dB)

van Haren Figure 5

69.5 70 70.5 71 71.5 72 72.5 73 73.5 74 74.5

−80−60−40−20 −5

0

5

Fig. 5. Overview of ADCP observations betweenz=7.8–86.8 m. In addition to the three current components cross-slope (u), along-slope (v)and bottom-normal (w) the corrected “error” velocity (e) and the relative echo intensity (dI ) are shown. The brown speckles throughout thegraphs indicate “bad data”, of which there are many.

Over the depth range between 420–580 m depth, which isslightly expanded with respect to the NIOZ2 depth rangeto cover the entire temperature range sampled by NIOZ2,δσθ=−0.108±0.002 δT kg m−3 (Fig. 3d). As salinity de-creased with depth it counteracted the temperature induceddensity gradient. Some of observed deviations from this lin-ear relationship were due to mismatches of the CTD sensorsthat were difficult to mend even in post-processing. The re-mainder of the mismatches was due to the environment thatvaried with time, with density steps and homogeneous layersalternating quickly. Although homogeneous layers of severalmeters in the vertical were abundant (Sect. 3.3.2), the varia-tions with time caused some problems for the in-situ calibra-tion of NIOZ2 (Appendix A).

3.2 General temperature and current observations near thebottom of GMS

3.2.1 General temperature observations

An overview of the entire record demonstrates the great vari-ability of temperature in the bottom boundary layer abovethe 4.5◦ slope near the top of GMS (Fig. 4a). This slopeis much less than the slope of∼1.6◦ of semidiurnal inter-nal tidal wave characteristics for typical interior stratifica-

tion and we do not expect internal tidal generation or criti-cal reflection at this site. Nevertheless, the main variabilityhas a semidiurnal tidal periodicity and a considerable dis-crepancy is observed across the vertical, as can be inferredfrom current meter temperature sensors atz=1.25 and 86 mthat span a larger range than NIOZ2. Away from the bot-tom many small-scale temperature fluctuations are superim-posed on the tidal signal. Very close to the bottom the recordis much smoother, but the non-linearity seems larger thanfurther up as is visible from the cnoidal shape of the tidalrecord, with occasional sharp ramps like near day 73.5. Indetailed NIOZ2 observations like in Figs. 4b and 4c it ap-pears that this ramp is extremely sharp, with a front passingthe sensors with temperature dropping by 0.5◦C within 1 s.Such fronts or bores have been observed previously, in datasampled at a rate of once/30 s above the continental slopesof the Faeroe-Shetland Channel, where they appeared at a 4days periodicity (Hosegood et al., 2004), and the Bay of Bis-cay, where tidal periodicity dominated (Gemmrich and vanHaren, 2001). Apparently, such roaring, upslope motions re-sulting in strongly non-linear temperature records also existat the sides of seamounts in the open ocean.

Preceding and following such frontal passage, near-bottom temperature stratification varied considerably, al-though completely homogeneous bottom boundary layers



−z

(m) T

(oC)

72.9 72.95 73 73.05 73.1 73.15 73.2 73.25 73.3 73.35 73.4−60

−40

−2012

13

−z

(m) −dT−−

dz(oC/m)

−60

−40

−20

−0.04

−0.02

0

0.02

−z

(m) u

−60

−40

−20−20

0

20

−z

(m) w

−60

−40

−20−2

0

2

−z

(m) e

−60

−40

−20−2

0

2

yearday

−z

(m) dI

(dB)

72.9 72.95 73 73.05 73.1 73.15 73.2 73.25 73.3 73.35 73.4

−60

−40

−20 −5

0

5

Fig. 6. A typical tidal period of ADCP and NIOZ2 observations betweenz=2.8–66.8 m. In∂T /∂z blue-green indicates stable stratification,yellow indicates neutral and orange-red indicates unstable overturning.

of several tens of meters thick as in flat-bottom tidal seaswere not observed. With the varying stratification manysmall-scale features were observed ranging from overturns(Sect. 3.3.3) to high-frequency internal waves (Sect. 3.3.4).The latter varied from waves having an apparent periodic-ity of 30–60 s, for example at day 73.62, and smooth 15-minwaves, which is of the order of the local buoyancy period,e.g. around day 73.7.

3.2.2 General current observations

Like temperature, the near-bottom currents and relative echointensity are also dominated by semidiurnal tidal variations(Fig. 5). As is obvious from the currents, the observa-tions start just after neaps and stop just before springs. Thecross-slope current generally has a slightly larger amplitudethan the along-slope tidal current, which has been mod-elled and observed before (Fig. 1; Mohn and Beckmann,2002). Apparently, the seamount does not steer the large-scale (∼“barotropic”) tidal current. These tidal currents arealso not uniform in the vertical, with strongest currents ob-served usually, but not always, not too far off the bottom(z∼10–20 m). Additionally, strong high-frequency (∼N )variability is observed, during which relatively large bottom-normal currents O(10−2 m s−1) are observed (the verticalstriping in thew panel). The lack of coherence betweene

andw, e being generally much smaller thanw, implies confi-dence in measuredw over scales larger than the beam spreadO(10 m), because this implies that the horizontal current in-homogeneity over the beam spread is much less than theactual bottom normal current (see definitions in Sect. 2.2).Each tidal period has its own characteristics, with varyingamplitudes and height of tidal and high-frequency motionsand echo intensity.

3.3 Detailed observations near the bottom of GMS

3.3.1 A tidal period

An example of a single tidal period (Fig. 6) shows the com-mon features, of which the details may vary during other tidalperiods, 1. asymmetry inu− (andv−, not shown) currentswith height, with near-bottom currents always leading thosehigher up, 2. a sudden transition to upslope motion and asmooth transition to downslope motion as in Hosegood etal. (2004), 3. many high-frequency waves in bothu andw,but not ine. Such waves exist up toz=80 m and they have pe-riods of typically∼2TN , TN denoting the large-scale buoy-ancy period, as observed previously in the North Sea (vanHaren et al., 2001), 4. these high-frequency motions are in-ternal waves as is evident from the high-resolution temper-ature measurements (especially from the panel on∂T /∂z,



13.33513.3413.34513.3513.3550

5

10

15

20

25

T (oC)

z (m

)

a

van Haren Figure 7

13.33513.3413.34513.3513.355

T (oC)

b

13.33513.3413.34513.3513.3550

5

10

15

20

25

T (oC)

z (m

)

c

Fig. 7. NIOZ2 sensor stability in near-homogeneous layers. In eachpanel colours indicate 10 NIOZ2 profiles, or 10 s of data, with inblack the local adiabatic lapse rate.(a) 40 s before panel(b). 5 sbefore−5 s after day 73.4868.(c) 40 a later than panel (b).

which demonstrate that a weakly stratified layer, not a homo-geneous bottom boundary layer, may extend up to z∼40 mlater in the up-slope phase of the tide, but also that stratifi-cation is capping a very thin bottom boundary layer of lessthan 5 m during the up- and down-slope phases. Especiallyduring the down-slope phase (between days 73.3–73.4)e isnon-negligible with respect tow, which is evidence of sub-stantial current inhomogeneity across the beam spread. Thiscurrent inhomogeneity seems to reflect small-scale motions,and during this period most overturns are observed in theNIOZ2 data, with the exception of the period when a tidalbore passed at the start of the upslope phase. A coarse esti-mate of Thorpe overturning scales during such period of en-hanced current inhomogeneity yieldsK=8±4·10−4 m2 s−1,whilst values are estimated between 3·10−4-3·10−2 m2 s−1,during brief periods at the start of the upslope phase.

3.3.2 Sensor stability: performance of NIOZ2 in near-homogeneous layers

Periods of very weak stability or near-homogeneous watersoffered the chance to investigate the stability of the NIOZ2sensors and to verify the accuracy obtained from the calibra-tion post-calibration adjustment (Appendix A). Such periodsoccurred above the sloping bottom of GMS, but generallyaway from the lowest sensors and for relatively brief peri-ods of time only. During a tidal period such periods usu-ally occurred twice, once well into the upslope phase, forexample around day 73.62 betweenz=20–40 m (not shown),and towards the very end of the downslope phase, for ex-ample around day 73.487 at several depths betweenz=8–25 m (Fig. 7). In both cases near-homogeneity lasted forabout 15-60 s at a particular depth. Near-homogeneity orneutral stability was assumed when a negative temperaturegradient was observed, which approached the adiabatic lapserate 0=−1.45±0.01·10−4◦

C m−1 in Fig. 7. The observedtemperature profile has a significant negative slope and ap-proaches0 betweenz=14–20 m in Fig. 7b andz=9–13 m in

−z

(m)

u (10−2 m s−1)

b

−50

−45

−40

−35 −20

−10

0

10

20yearday

−z

(m)a

73.46 73.47−50

−45

−40

−35

−30

−25

−20

−15

−10

−5

yearday

−z

(m)

1 min

T (oC)

c

van Haren Figure 8

73.462 73.464 73.466 73.468

−50

−45

−40

−35 13.44

13.45

13.46

13.47

13.48

13.49

13.5

13.51

Fig. 8. Example of detailed observation of an overturn,∼12 mbreaking wave aroundz=40 m during the downslope phase of thetide. (a) Corrected data showing temperature stratification and veryhigh-frequency (∼1 min) temperature variations down to the sen-sors closest to the bottom.(b) Detailed period (indicated by rectan-gle in (a)) of ADCP’s u-current.(c) From the white rectangle in (a)showing a large Kelvin-Helmholtz overturning (Turner, 1973).

Fig. 7c, or across∼10 independent sensors each time. Thissuggests a much greater stability of∼10−4◦

C of the NIOZ2sensors, and the relatively poor accuracy is entirely due to thecalibration procedure, so that improvement is suggested formanual adjustment of the calibration (Appendix A). Furtherexamples of detailed observations are given after applicationof such adjustment to the calibration using the sensors’ sta-bility.

3.3.3 Temperature overturns

NIOZ2 revealed the character of some of the occasionalsmall-scale temperature overturns during the downslopephase of the tide (Fig. 8). Such overturns appeared in a moreor less regular sinusoidal waveform in a temperature time se-ries, but they were related to a breaking wave or rolling-upof a Kelvin-Helmholtz overturn (Figs. 8a and 8c). Althoughsuch overturning waves have been observed in the ocean atshallow depths using photography (Woods, 1968), laboratoryexperiments (examples in Turner, 1973), temperature profil-ing (Thorpe and Hall, 1974) and detailed acoustic backscat-ter (Orr and Mignerey, 2003; Moum et al., 2003), the presentobservations are the first detailed observations of such over-turn at great depth. As the overturn is poorly resolved and notrecognizable in the ADCP data (Fig. 8b), which are sampled“only” once per 30 s resolving “only” O(10 m) horizontallyand 1 m vertically, such NIOZ2 observations emphasize theneed to sample at high frequency to capture such processes.

The large-scale overturn, backwards breaking as in Turner(1973), of more than 10 m in height, and the associated verythin layers of enhanced stratification of∼1 m, take about2 min to pass the sensors. Several are observed in Fig. 8a, be-



12.2

12.3

12.4

T (

o C)

yearday

a

yearday

−z

(m)

1 min T(oC)

b

van Haren Figure 9

73.616 73.618 73.62 73.622 73.624

−50

−45

−40

−35

12.15

12.2

12.25

12.3

12.35

12.4

Fig. 9. Example of NIOZ2 observations of very high-frequency in-ternal “wave” during the upslope phase of the tide.(a) temperaturefrom 5 sensors at 1 m intervals between the white dashed lines in(b). Contour plot using information of all sensors in the depth-timeinterval shown.

tween 10–20 m around day 73.458 and also in the white rect-angle, which is Fig. 8c. As u∼0.1±0.05 m s−1 and assum-ing the phase speedc∼u, the horizontal length scale of theoverturn is∼10 m. These horizontal length scale estimatesare half an order of magnitude smaller than those observedby Orr and Mignerey (2003) and Moum et al. (2003), whilstthe vertical length scales are similar. The latter authors alsoobserved intensification of acoustic backscatter in layers of∼0.5–1 m thickness, suggesting similar overturning in theselayers as well. Such scales are barely resolved by the presentobservations, but small-scale variations O(1–10 s), O(1–5 m)are visible in the NIOZ2 observations (Fig. 8c).

3.3.4 (Too) short internal waves

The above fast sampling is not only required for large-scale‘turbulent’ motions, but also for small-scale waves. NIOZ2also revealed occasional “regular” small-scale temperaturevariations (Fig. 9) that appeared to have periods of∼3 min,much less than the local buoyancy period of 20–30 min(N=50–70 cpd) (Fig. 3c), no matter how small the verticallength scale was chosen to computeN(z). Figure 9 showsthe instrumental capability in the occasionally very smoothtemperature records hiding the 0.5–1·10−3◦

C noise and theoccasionally quite “noisy” environmental background, withnoise levels of O(10−2◦

C), especially also during the passageof a wave (e.g. between days 73.6185–73.62). The smoothwaves have a period of∼3 min, whilst the noisy motionsare also periodic (e.g. around day 73.621, with period of∼10 s). These smooth waves are tilted at an angle to thevertical. If such waves existed as free propagating inter-nal gravity waves their upward “phase” propagation wouldimply downward energy propagation. However, their ob-

74.511 74.5115 74.512 74.5125 74.513

13.4

13.42

13.44

13.46

13.48

13.5

13.52

T (

o C)

yearday

1 min

b

van Haren Figure 10

yearday

−z

(m)

T(oC)

a

74.51 74.515 74.52 74.525−50

−40

−30

−20

−10

13.1

13.2

13.3

13.4

13.5

13.6

13.7

Fig. 10. Example of NIOZ2 observations of very high-frequencyinternal “wave” and occasionally unstable fluid during the upslopephase of the tide.(a) Contour plot showing lower temperature wa-ter (green; between days 74.51–74.514; 38<z<50 m) in an environ-ment of warmer water under the crest of a large-scale wave, whichhas a period near the local buoyancy period. The dashed black linesof the small-scale “wave” indicate the depth levels and time rangeof the temperature series in(b).

served frequencyσo�N . Computing the intrinsic frequencyσ=N=σo−U · k, U≈0.15±0.05 m s−1 andk the horizontalwavenumber, we require for free internal waves a wavelengthλ=2π/k≈2πU/σo=27±9 m. Apparently the short waves aregenerated and Doppler shifted by the larger scale internalwaves. They appear as interfacial waves and they may gen-erate turbulence as they “move” at the fringe of breaking.

This can also be seen in a second example of even smallerscale smooth waves, which have periods of less than 1 min,and which are “carried” below the crest of a wave havingan ∼18 min period, close to the buoyancy period (Fig. 10).NIOZ2 is capable of resolving both the 30 s “wave”, as wellas coherent and incoherent motions within that small wave,which both exist as cool water intrusions under the carrierwave. It may be obvious that these small-scale waves arenot visible in the ADCP data, not just because they are notresolved in time, but especially also because they have hori-zontal length scales that are comparable to beam spread overwhich currents are averaged. Resolving currents for thesephenomena thus requires other current devices than an acous-tic profiler.

4 Summary

We have presented some data from a newly built thermis-tor string, which is capable of measuring temperature varia-tions in the deep ocean at an estimated precision better than1.5·10−3◦

C, whilst sampling 128 sensors synchronously ata rate of 1 Hz for a period of up to 15 days potentially at



74.8 74.81 74.82 74.83 74.8410

15

20

O

OT (

o C)

a

74.8024 74.8026 74.8028 74.803 74.8032 74.803410.651

10.652

10.653

10.654

10.655

10.656

T (

o C)

yearday

b

74.8256 74.8258 74.826 74.8262 74.8264 74.8266

14.098

14.1

14.102

14.104

14.106

T (

o C)

c

Fig. 11. In situ calibration of NIOZ2 using CTD.(a) CalibratedNIOZ2 temperatures from 10 sensors (colours), and upcast trace ofCTD (black) demonstrating the 9 steps between 10.5–18.8◦C whenthe CTD is held still in a layer of “constant” temperature.(b),(c) (different scales) Noise or precision comparison with environ-mental small-scale variability of 86 s of the record in (a) (circles),which demonstrate the 3 times higher noise rate of the NIOZ2 sen-sors compared to the CTD-temperature sensor, but also some of therelatively large environmental variability in∼20 s waves.

depths down to 6000 m. During the construction and firsttrials the sensors turned out very robust, but a major prob-lem was found in the proper bonding of polyurethane castingresin to titanium.

The response and sampling time in combination with itsendurance have opened new capabilities for studies on inter-nal waves, e.g. near sloping bottoms, up to the highest fre-quencies, including non-linearities like internal bores. Thelatter appear irregularly at a relatively large time scale of sev-eral hours–days, but pass the sensors within a few minutes.This is not different at GMS, where the tidal variability seemsdominant, but the amplitude of such bores is modulated witha much longer timescale. As the importance of internal waveinduced mixing is more and more recognized, also in con-junction with large-scale ocean circulation, detailed studiesare needed to learn more about such mixing processes. Pre-viously, detailed studies have been performed in laboratoryexperiments (e.g. McEwan, 1983) or near the surface of theocean (e.g. Woods, 1968; Marmorino et al., 1987; Moum etal., 1992), but very limited in the deep ocean (Thorpe, 1987).Especially the transfer of energy from the large scale to thesmall mixing scales was difficult to explore with deep-oceanobservations. In the examples given in this paper clearly theDoppler-shift of small internal waves to the point of breakingis observed, as well as shear-induced Kelvin-Helmholtz over-turning. The appearance of bores during the upslope phaseof the tide remains a subject of study, especially also becausewe have not yet established the underlying reason why themagnitude of a bore varies so strongly between the differ-ent (tidal) periods. Future investigations will also focus on

10 12 14 16 18

6.5

7

7.5

8

8.5

9

x 106

T (oC)

elec

tron

ic v

alue

a

−2 −1 0 1 2

x 10−3

10

12

14

16

18

T (

o C)

Tfit−T

obs (oC)

b

van Haren Figure 12

Fig. 12. (a)Polynomial fit of NIOZ2 raw data to 7 CTD temperaturevalues following in situ calibration.(b) Residual of the fit (withmean std=1.3·10−3◦

C).

small-scale fronts and homogeneous layers as resolved byNIOZ2.

Appendix A: Calibration of NIOZ2

As the digital representations of the data are non-linearly re-lated to temperature, noise level is also a function of temper-ature, ranging from 4·10−4◦

C at 0◦C to 1·10−3◦C at 40◦C.

Due to this non-linearity a precise calibration of each of thethermistors is required to achieve an accuracy of<5·10−3◦

C.A “best achievable” laboratory calibration takes months andis a complicated activity, whilst the accuracy is not betterthan 3·10−3◦

C (van Haren et al., 2001). As a result, likeNIOZ1, NIOZ2 is calibrated in situ immediately before andafter a deployment using a high-performance CTD: a muchfaster calibration.

The thermistor string is coiled up on top of the Rosette-sampler protective cage surrounding a Seabird-911 plusCTD. The temperature sensor of the CTD is accurate towithin 1·10−3◦

C, with a polynomial deviation of<1·10−4◦C

and initial stability of 2·10−4◦C that can vary up to 2·10−4◦

Cper month. The distance between the CTD-temperature andpressure sensors and the thermistor string is about 1 m. Dur-ing the upcast the CTD is commanded to depths of layersof near-homogeneous water, which are selected during thedowncast (Fig. 11). As these layers move constantly up- anddown, or are modified by internal wave straining and mix-ing, the requirement is that the temperature varies less than1·10−3◦

C for a period of at least 90 s. Therefore, CTD’stemperature and pressure values are constantly monitoredvisually, which is not always always achieved satisfactorily(Fig. 11). Flushing of water passed the coiled up thermistorstring seems reasonable as we did not see much trend in therecord of individual sensors, and in comparison with thoseof other sensors, greater than the instrumental noise level of∼7·10−4◦

C at a particular calibration level. Given the de-



gree of non-linearity of the sensors 6–8 calibration values aresought, in this case over a range between 10.6–18.8◦C. Theelectronic NIOZ2 data are translated to temperatures using athird order polynomial fit to the calibration values for eachsensors separately (Fig. 12). The standard deviation of thispolynomial fit was between 1.2-1.5·10−3◦

C, resulting in anoverall accuracy of∼5·10−3◦

C. The above procedure takesabout 3 h for calibration around 2000 m depth.

In general, the result of this in-situ calibration is not muchbetter than the laboratory calibration, due to the environ-mental variability, mooring frame vibrations and flow ob-structions. This may also be inferred from the variabilityin CTD temperature, with standard deviations varying be-tween 2-8·10−4◦

C, all larger than the instruments’ stability,at the different calibration levels (Fig. 11). However, becausethe NIOZ2 sensors are much more stable (see example inSect. 3.3.2), post-processing can provide an accuracy closeto the 1·10−3◦

C precision (Fig. 13). This requires two moreprocessing steps after calibration.

First, profiles of mean temperatures are computed oversmall periods of time (∼TN , the local large-scale buoyancyperiod) in which temperature is reasonably constant. Theseprofiles are manually corrected to a new mean of a static sta-ble temperature profile (Fig. 13a) under the assumption thatoverturns and small-scale instabilities are removed when av-eraged over such period. In practice, the corrections are con-stant shifts in temperature (Figs. 13b and 13d) and generally<5·10−3◦

C (Fig. 13a), confirming the calibration accuracy.They are attributed to the difficulty of holding the calibrationCTD long enough in constant temperature waters (Fig. 11).This is concluded as this correction is valid each time theparticular correction temperature is reached during the en-tire period of observations. This first step is sufficient foran accuracy to within 1·10−3◦

C over a short period of time(Figs. 13c and 13e). However, when this accuracy is requiredfor the entire record, that is for the entire temperature rangeobserved, a second post-calibration step can be used that alsoaccommodates for the non-linearity of the sensors.

Over the entire temperature range of the record some 6–10 correction profiles as in Fig. 13 are then constructed. Thecalibrated data record is corrected by linear temporal inter-polation and replacing the means by the static stable meansfor each particular temperature in each sensor.

Acknowledgements.We thank the crew of the R. V. Pelagia fordeploying the “mixBB” bottom lander. We thank participants of theyellow tape-team, who taped NIOZ2 several times to its strengthmember: T. Hillebrand, K. Veth and P. Hosegood. M. Hiehle com-posed Fig. 1. The development of the NIOZ thermistor strings andthe deployment in the Canary Basin were financed by investmentgrants (“Oceanographic equipment” and “LOCO”, respectively)from the Netherlands organisation for the advancement of scientificresearch, NWO.

Edited by: M. Tomczak

13.46 13.48 13.535

40

45

50

T (oC)

z (m

)

a

van Haren Figure 13

b

−50

−45

−40

−35

T(oC)5 min

c

13.46

13.47

13.48

13.49

13.5

13.51

73.465 73.47

yearday

d

73.465 73.47

13.47

13.48

13.49

13.5

13.51

yearday

T (

o C)

e

Fig. 13. Example of post-calibration processing of NIOZ2.(a)Mean profile for 32 sensors of the data portion in (b): uncor-rected in red and “corrected” in black. The horizontal bar indicates±2 std=±4·10−3◦

C, 1 std denotes the standard deviation betweenthe profiles.(b) Uncorrected 22 min of data.(c) As (b), but replac-ing the observed mean with the corrected from (a).(d) Four tem-perature traces from (b) observed betweenz=45–47 m.(e) As (d),but corrected, demonstrating that the trace atz=46 m is no longerdisplaced with respect to the others.

References

Alford, M. H. and Pinkel, R.: Observations of overturning in thethermocline: the context of ocean mixing, J. Phys. Oceanogr.,30, 805–832, 2000.

Brandt, P., Rubino, A., Quadfasel, D., Alpers, W., Selschopp, J.,and Fiekas, H.-V.: Evidence for the influence of Atlantic-Ionianstream fluctuations on the tidally induced internal dynamics inthe Strait of Messina, J. Phys. Oceanogr., 29, 1071–1080, 1999.

Gemmrich, J. R. and van Haren, H.: Thermal fronts generated byinternal waves propagating obliquely along the continental slope,J. Phys. Oceanogr., 31, 649–655, 2001.

Hosegood, P., Bonnin, J., and van Haren, H.: Solibore-induced sed-iment resuspension in the Faeroe-Shetland Channel, Geophys.Res. Lett., 31, L09301, doi:10.1029/2004GL019544, 2004.

McEwan, A. D.: The kinematics of stratified mixing through inter-nal wavebreaking, J. Fluid Mech., 128, 47–57, 1983.

Marmorino, G. O., Rosenblum, L. J., and Trump, C. L.: Fine-scaletemperature variability: the influence of near-inertial waves, J.Geophys. Res., 92, 13 049–13 062, 1987.

Mohn, C. and Beckmann, A.: The upper ocean circulation at GreatMeteor Seamount, Part I: structure of density and flow fields,Ocean Dyn., 52, 179–193, 2002.

Moum, J. N., Hebert, D., Paulson, C. A., and Caldwell, D. R.:Turbulence and internal waves at the equator, Part I: statisticsfrom towed thermistors and a microstructure profiler, J. Phys.Oceanogr., 22, 1330–1345, 1992.

Moum, J. N., Farmer, D. M., Smyth, W. D., Armi, L., and Vagle,S.: Structure and generation of turbulence at interfaces strainedby internal solitary waves propagating shoreward over the conti-nental shelf, J. Phys. Oceanogr., 33, 2093–2112, 2003.



Orr, M. H. and Mignerey, P. C.: Nonlinear internal waves in theSouth China Sea: observation of the conversion of depressioninternal waves to elevation internal waves, J. Geophys. Res., 108,3064, doi:10.1029/2001JC00163, 2003.

Pinkel, R., Sherman, J. Smith, J., and Anderson, S.: Strain: observa-tions of the vertical gradient of isopycnal vertical displacement,J. Phys. Oceanogr., 21, 527–540, 1991.

Selschopp, J.: A towed CTD chain for high-resolution hydrography,Deep-Sea Res. I, 44, 147–165, 1997.

Thorpe, S. A.: Current and temperature variability on the continen-tal slope, Phil. Trans. R. Soc. Lond., A 323, 471–517, 1987.

Thorpe, S. A. and Hall, A. J.: Evidence of Kelvin-Helmholtz bil-lows in Loch Ness, Limnol. Oceanogr., 19, 973–976, 1974.

Turner, J. S.: Buoyancy Effects in Fluids, Cambridge UniversityPress, Cambridge, 1973.

van Haren, H., Oakey, N., and Garrett, C.: Measurements of internalwave band eddy fluxes above a sloping bottom, J. Mar. Res., 52,909–946, 1994.

van Haren, H., Groenewegen, R., Laan, M., and Koster, B.: A fastand accurate thermistor string, J. Atmos. Oceanic Technol., 18,256–265, 2001.

Woods, J. D.: Wave-induced shear instability in the summer ther-mocline, J. Fluid Mech, 32, 791–800, 1968.


Documents

High sampling rate thermistor string observations at the slope of …tmp.hansvanharen.nl/pedef/os-1-17.pdf · 2017-01-11 · the R-parts of a certain type of RC-oscillator, a Wien(bridge)